Dynamic risk assessment method for highway in heavy fog weather

Abstract

The invention provides a highway heavy fog weather dynamic risk assessment method, and relates to the technical field of traffic safety, and the method comprises the steps: collecting highway historical traffic data, and extracting and preprocessing a historical risk factor data set to construct a heavy fog weather risk analysis database; analyzing the incidence relation between the risk factor and the accident, extracting the space-time and dynamic characteristics of the risk factor, identifying a combined action mode, and building a four-layer risk fusion assessment framework after obtaining a risk evolution rule; and inputting current risk factor data, carrying out framework processing, and outputting a quantitative dynamic risk assessment value with a confidence interval. According to the method, the staged adjustment of risk modeling is realized by extracting the multi-dimensional dynamic features, the weight is dynamically corrected, the factor interaction is quantified, the evaluation value is iteratively optimized, and the confidence interval is output, so that the limitation that the traditional static modeling cannot adapt to the dynamic change of the heavy fog risk is broken through, and the real-time performance, the accuracy and the scene adaptability of risk evaluation are greatly improved.

Description

Technical Field

[0001] This invention relates to the field of traffic safety technology, and more specifically, to a method for dynamic risk assessment of heavy fog on highways. Background Technology

[0002] With the increasing coverage of highway networks and the year-on-year growth of daily traffic volume, foggy weather, due to its significant reduction in atmospheric visibility and weakening of road surface friction coefficient, has become a core contributing factor to traffic accidents such as multi-vehicle collisions and traffic congestion. Meanwhile, the development of intelligent transportation technologies has accumulated massive amounts of historical traffic data, leading to an increasingly strong demand from the industry for dynamic early warning and precise control of risks under foggy weather conditions. However, existing technologies struggle to fully extract the spatiotemporal patterns of risk factors from the data and cannot adapt to the dynamic changes in fog risks over time and road sections. Consequently, traffic safety management lacks efficient risk assessment technology support.

[0003] Traditional methods for assessing the risk of heavy fog on highways primarily rely on static modeling, using risk factor data from fixed areas and single time periods. They fail to extract the spatial distribution characteristics, temporal fluctuations, and dynamic trends of fog throughout its entire lifecycle (formation, development, and dissipation) from historical data. Weight calculations depend on subjective experience and do not dynamically adjust the judgment matrix based on the correlation and evolution of risk factors and accidents. Furthermore, they ignore the synergistic / antagonistic effects of risk factors, fail to quantify the interactive impacts of factors, and produce assessment results that are merely single values ​​without confidence intervals. These methods cannot adapt to the dynamic changes in fog risk, resulting in poor real-time performance, low accuracy, and difficulty in supporting dynamic traffic safety management.

[0004] Therefore, it is necessary to design a dynamic risk assessment method for heavy fog on highways to address the problems of existing technologies that use static modeling, which cannot extract dynamic characteristics of the fog stage, dynamically adjust weights and quantify the interaction relationship of factors, and have no confidence interval for the assessment results, which cannot adapt to the dynamic changes of fog risk, resulting in insufficient real-time performance, accuracy and scenario adaptability of the assessment. Summary of the Invention

[0005] In view of this, the present invention proposes a dynamic risk assessment method for heavy fog weather on highways, which aims to solve the problems of existing technologies that use static modeling, cannot extract dynamic features of the fog stage, dynamically correct weights and quantify the interaction relationship of factors, have no confidence interval for assessment results, and cannot adapt to the dynamic changes of fog risk, resulting in insufficient real-time performance, accuracy and scenario adaptability of the assessment.

[0006] In one aspect, this invention proposes a dynamic risk assessment method for heavy fog weather on highways, comprising: Historical traffic data of highways are collected, and historical risk factor datasets are extracted from the historical traffic data. The historical risk factor datasets are preprocessed, and a fog weather risk analysis database is constructed based on the preprocessed historical risk factor datasets. The correlation between risk factors and the probability of accident occurrence is obtained, the spatial distribution characteristics, temporal variation characteristics and dynamic variation trends of historical risk factor data are extracted, and the combined effect patterns among the historical risk factor data are identified. Based on the spatial distribution characteristics, temporal variation characteristics, dynamic change trends, and combined action modes, the evolutionary characteristics and patterns of risk are obtained; based on the evolutionary characteristics and patterns, a risk fusion assessment framework is constructed, and the risk fusion assessment framework is divided into a risk factor input layer, a weight analysis layer, a fusion calculation layer, and a risk output layer; The current risk factor data is obtained, and the current risk factor data is processed sequentially through the risk factor input layer, the weight parsing layer, the fusion operation layer, and the risk output layer, and a quantitative dynamic risk assessment value with confidence interval is output. The risk factor input layer preprocesses the current risk factor data; the weight parsing layer obtains the basic weights of the current risk factor data based on the historical risk factor data and the judgment matrix; the fusion operation layer obtains the multi-factor joint fuzzy measure of the current risk factor data based on the fuzzy integral fusion model; and the risk output layer obtains and outputs the quantitative dynamic risk assessment value based on the evolutionary features and laws and the multi-factor joint fuzzy measure.

[0007] Furthermore, when obtaining the correlation between risk factors and the probability of an accident occurring, this includes: The historical risk factor dataset includes several historical risk factor data, including: atmospheric visibility, road surface friction coefficient, and traffic flow density. The first atmospheric visibility dataset is obtained by extracting the historical risk factor data corresponding to the maximum atmospheric visibility from the historical risk factor dataset, and the second atmospheric visibility dataset is obtained by extracting the historical risk factor data corresponding to the minimum atmospheric visibility from the historical risk factor dataset. The atmospheric visibility variation range is obtained by subtracting the atmospheric visibility minimum value from the maximum atmospheric visibility value. The atmospheric visibility accident probability variation range is obtained by subtracting the accident probability from the accident probability in the first atmospheric visibility dataset. The atmospheric visibility accident probability variation range is obtained by dividing the atmospheric visibility variation range by the atmospheric visibility accident probability variation range. The first road surface friction coefficient dataset is obtained by extracting the historical risk factor data corresponding to the maximum value of the road surface friction coefficient from the historical risk factor dataset, and the second road surface friction coefficient dataset is obtained by extracting the historical risk factor data corresponding to the minimum value of the road surface friction coefficient from the historical risk factor dataset. The change range of the road friction coefficient is obtained by subtracting the minimum value of the road friction coefficient from the maximum value of the road friction coefficient; the change range of the road friction coefficient accident probability is obtained by subtracting the accident probability of the second road friction coefficient dataset from the accident probability of the first road friction coefficient dataset; and the road friction coefficient correlation ratio is obtained by dividing the change range of the road friction coefficient by the change range of the road friction coefficient accident probability. The first traffic flow density dataset is obtained by extracting the historical risk factor data corresponding to the maximum traffic flow density from the historical risk factor dataset, and the second traffic flow density dataset is obtained by extracting the historical risk factor data corresponding to the minimum traffic flow density from the historical risk factor dataset. The traffic flow density change range is obtained by subtracting the minimum traffic flow density from the maximum traffic flow density. The traffic flow density accident probability change range is obtained by subtracting the accident probability from the accident probability in the first traffic flow density dataset. The traffic flow density correlation ratio is obtained by dividing the traffic flow density change range by the traffic flow density accident probability change range.

[0008] Furthermore, when obtaining the correlation between risk factors and the probability of an accident, it also includes: When the atmospheric visibility correlation ratio is greater than the correlation ratio threshold, the correlation between atmospheric visibility and the probability of an accident is determined to be a strong correlation, and atmospheric visibility is marked as a core risk factor; when the atmospheric visibility correlation ratio is less than or equal to the correlation ratio threshold, the correlation between atmospheric visibility and the probability of an accident is determined to be a weak correlation, and atmospheric visibility is marked as a secondary risk factor. When the correlation ratio of the road surface friction coefficient is greater than the correlation ratio threshold, the correlation between the road surface friction coefficient and the probability of accident occurrence is determined to be a strong correlation, and the road surface friction coefficient is marked as a core risk factor; when the correlation ratio of the road surface friction coefficient is less than or equal to the correlation ratio threshold, the correlation between the road surface friction coefficient and the probability of accident occurrence is determined to be a weak correlation, and the road surface friction coefficient is marked as a secondary risk factor. When the traffic flow density correlation ratio is greater than the correlation ratio threshold, the correlation between the traffic flow density and the probability of an accident is determined to be strong, and the traffic flow density is marked as a core risk factor; when the traffic flow density correlation ratio is less than or equal to the correlation ratio threshold, the correlation between the traffic flow density and the probability of an accident is determined to be weak, and the traffic flow density is marked as a secondary risk factor.

[0009] Furthermore, when extracting the spatial distribution characteristics, temporal variation characteristics, and dynamic trends of historical risk factor data, and identifying the combined interaction patterns among the historical risk factor data, the process includes: The historical risk factor dataset is divided into subsets of different regions according to spatial road segments. The mean values ​​of each subset are calculated at different time points, and the changes in the mean values ​​reflect the spatial distribution characteristics. The numerical fluctuation values ​​of the same subset are calculated over a continuous period of time, and the numerical fluctuation values ​​reflect the temporal change characteristics. The numerical change curves of the historical risk factor data during the formation, development, and dissipation stages of fog are fitted using time series analysis methods, and the dynamic change trends are extracted by the slope of the numerical change curves. The core risk factor is combined with the secondary risk factor to obtain multiple risk factor combinations, and each risk factor combination contains at least two individual risk factors with different relationships. The historical risk factor data of the accidents that occurred are extracted from the historical risk factor dataset to obtain the accident dataset; The risk factor combination is extracted from the accident dataset to obtain the accident combination dataset, and the risk factor combination in the accident combination dataset is the accident risk factor combination; the risk factor combination is extracted from the historical risk factor dataset to obtain the overall combination dataset; the proportion of the accident risk factor combination in the overall combination dataset is the probability of occurrence of the risk factor combination accident; The individual risk factor is extracted from the accident dataset to obtain a single risk factor dataset, and the individual risk factor in the accident combination dataset is a single accident risk factor; the individual risk factor is extracted from the historical risk factor dataset to obtain an overall single risk factor dataset; the proportion of the individual accident risk factor in the overall single risk factor dataset is the probability of occurrence of the single risk factor accident.

[0010] Furthermore, when extracting the spatial distribution characteristics, temporal variation characteristics, and dynamic trends of historical risk factor data, and identifying the combined interaction patterns among the historical risk factor data, the process also includes: When the probability of occurrence of the combined risk factor event is greater than or equal to the sum of the probabilities of occurrence of each individual risk factor event, the combined action mode of the current risk factor combination is a synergistic action mode. When the probability of an event occurring in the combination of risk factors is less than the sum of the probabilities of each individual risk factor event, the combined action mode of the current risk factor combination is an antagonistic action mode.

[0011] Furthermore, when obtaining the evolutionary characteristics and patterns of risk based on the spatial distribution characteristics, the temporal variation characteristics, the dynamic change trend, and the combined action mode, it includes: Based on the spatial distribution characteristics, the correspondence between the mean values ​​of the subsets in different regions and the frequency of accidents is statistically analyzed. When the change in the mean value is greater than a first threshold, the risk in the current region is more affected by the synergistic effect mode than by the antagonistic effect mode. When the change in the mean value is less than or equal to the first threshold, the risk in the current region is more affected by the antagonistic effect mode than by the synergistic effect mode. Based on the time change characteristics, the correspondence between the numerical fluctuation values ​​of the same subset of data and the time period of the accident is statistically analyzed. When the numerical fluctuation value is greater than the fluctuation value threshold, the risk in the current period fluctuates rapidly with the change in the value of the core risk factor; when the numerical fluctuation value is less than or equal to the fluctuation value threshold, the risk in the current period remains stable. Based on the dynamic change trend and the slope of the numerical change curve, the risk change trend at different stages of fog is summarized. When the fog is in the formation stage, the risk increases with the increase of the slope; when the fog is in the development stage, the risk stabilizes as the slope approaches zero; and when the fog is in the dissipation stage, the risk decreases as the slope decreases. Based on the correlation between the numerical changes of the risk factors and the combined action patterns, the evolution law of risk with historical risk factor data is summarized. When the value of the core risk factor is in the extreme range and the core risk factor and the secondary risk factor form a synergistic action pattern, the risk changes exponentially; when the value of the core risk factor is in the normal range and the core risk factor and the secondary risk factor form the antagonistic action pattern, the risk changes linearly.

[0012] Furthermore, when constructing a risk fusion assessment framework based on the aforementioned evolutionary characteristics and patterns, it includes: A regionalized data processing submodule is set up in the risk fusion assessment framework, and risk factor data for different road sections are processed differently. A phased calculation submodule is set up in the risk fusion assessment framework, and the risk calculation logic is matched to the formation, development and dissipation stages of fog. A coupling effect quantification submodule is set up in the framework to quantify the interaction relationship between the core risk factor and the secondary risk factor under the synergistic and antagonistic modes.

[0013] Furthermore, when the weight parsing layer obtains the basic weights of the current risk factor data based on the historical risk factor data and the judgment matrix, it includes: An initial judgment matrix is ​​constructed based on the risk factor types of the current risk factor data. The rows and columns of the initial judgment matrix correspond to the risk factor types of the current risk factor data, and the matrix elements are assigned values ​​to represent the relative importance of different current risk factor data. The correlation results corresponding to the historical risk factor data and the evolution law of risk with the historical risk factor data are retrieved from the fog weather risk analysis database. The correlation results and the evolution law are used together as correction coefficients to adjust the matrix elements of the initial judgment matrix to obtain the corrected judgment matrix. The consistency ratio of the corrected judgment matrix is ​​calculated. When the consistency ratio is greater than the consistency threshold, the matrix elements of the initial judgment matrix are readjusted until the consistency ratio is less than or equal to the consistency threshold. When the consistency ratio is less than or equal to the consistency threshold, the current risk factor data is normalized. For negative factor data such as atmospheric visibility, the reciprocal standardization method is used. For positive factor data such as road friction coefficient and traffic flow density, the extreme value method is used to map to a unified numerical range. The normalized current risk factor data is substituted into the corrected judgment matrix. The maximum eigenvalue and corresponding eigenvector of the corrected judgment matrix are calculated using the eigenvalue method. The eigenvector is normalized to obtain the basic weight of the current risk factor data. The weight adjustment rules for different stages of fog are matched through the staged operation submodule, and the adjusted basic weight is passed to the fusion operation layer.

[0014] Furthermore, when the fusion operation layer obtains the multi-factor joint fuzzy measure of the current risk factor data based on the fuzzy integral fusion model, it includes: The coupling effect quantification submodule uses the adjusted base weight values ​​and the evolution pattern of risk with the historical risk factor data to determine the single-factor fuzzy measure value of each current risk factor data. The single-factor fuzzy measure value is positively correlated with the base weight values. The fluctuation range of the single-factor fuzzy measure value is calculated. When the fluctuation range of the measure value is greater than the second fluctuation threshold, the single-factor fuzzy measure value is smoothed. When the fluctuation range of the measure value is less than or equal to the second fluctuation threshold, a multi-factor joint fuzzy measure is calculated based on the single-factor fuzzy measure value. When calculating the multi-factor joint fuzzy measure, the synthesis rules of fuzzy integrals are combined with the synergistic and antagonistic effects to perform a weighted summation on multiple single-factor fuzzy measure values. The weight coefficients of the weighted summation are adjusted based on the combined effect mode of the risk factor combination. When the combined effect mode is the synergistic effect mode, the corresponding weight coefficient is increased; when the combined effect mode is the antagonistic effect mode, the corresponding weight coefficient is decreased. After the calculation is completed, the multi-factor joint fuzzy measure is passed to the risk output layer.

[0015] Furthermore, when the risk output layer obtains and outputs the quantitative dynamic risk assessment value based on the evolutionary characteristics and patterns and the multi-factor joint fuzzy measure, it includes: Based on the evolution characteristics of risk with spatial road segments and time dimensions, as well as the evolution law of risk with the historical risk factor data, a corresponding nonlinear calculation method is selected, and the multi-factor joint fuzzy measure is substituted into the nonlinear calculation method for initial calculation to obtain a preliminary risk assessment value. The current risk factor data is incorporated into the nonlinear calculation method in real time for iterative optimization. The iterative parameters are adjusted by combining the dynamic changes in the formation, development and dissipation stages of the fog through the staged calculation submodule, and the deviation of the preliminary risk assessment value before and after iterative optimization is calculated. When the deviation value is greater than the deviation threshold, new current risk factor data is continuously incorporated and the iteration parameters are adjusted in conjunction with the spatial distribution characteristics until the deviation value is less than or equal to the deviation threshold; when the deviation value is less than or equal to the deviation threshold, the iteration is stopped and the confidence interval of the preliminary risk assessment value is calculated. When calculating the confidence interval, the standard deviation is calculated using the numerical distribution of the iteratively optimized risk assessment value. The weight of the standard deviation is adjusted in combination with the risk fluctuation characteristics of different regions in the spatial distribution characteristics. The upper and lower limits of the confidence interval are determined based on the adjusted standard deviation. The iteratively optimized risk assessment value is combined with the confidence interval to form the quantitative dynamic risk assessment value with confidence interval and output.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: The dynamic risk assessment method for heavy fog on highways is based on historical risk factor data, extracting spatial distribution, temporal changes, and dynamic characteristics of fog stages. It matches the risk calculation logic of fog formation, development, and dissipation stages through a phased computational submodule, achieving phased dynamic adjustment of risk modeling. The weight analysis layer dynamically corrects the judgment matrix and adjusts factor weights based on correlation and evolutionary patterns. The fusion computation layer dynamically quantifies the interaction relationship between core and secondary risk factors based on synergistic / antagonistic patterns. The risk output layer iteratively optimizes by incorporating current risk factor data in real time, dynamically corrects the assessment value, and outputs it in conjunction with confidence intervals. This overcomes the limitations of traditional static modeling in adapting to dynamic changes in fog risk, significantly improving the real-time performance, accuracy, and scenario adaptability of risk assessment. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a dynamic risk assessment method for heavy fog on highways provided in an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the invention to those skilled in the art. It should be noted that, without conflict, the embodiments and features described herein can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] Reference Figure 1 As shown in some embodiments of this application, a dynamic risk assessment method for highway fog weather includes: Historical traffic data of highways are collected, and historical risk factor datasets are extracted from the historical traffic data. The historical risk factor datasets are preprocessed, and a fog weather risk analysis database is constructed based on the preprocessed historical risk factor datasets. The correlation between risk factors and the probability of accident occurrence is obtained, the spatial distribution characteristics, temporal variation characteristics and dynamic variation trends of historical risk factor data are extracted, and the combined effect patterns among the historical risk factor data are identified. Based on the spatial distribution characteristics, temporal variation characteristics, dynamic change trends, and combined action modes, the evolutionary characteristics and patterns of risk are obtained; based on the evolutionary characteristics and patterns, a risk fusion assessment framework is constructed, and the risk fusion assessment framework is divided into a risk factor input layer, a weight analysis layer, a fusion calculation layer, and a risk output layer; The current risk factor data is obtained, and the current risk factor data is processed sequentially through the risk factor input layer, the weight parsing layer, the fusion operation layer, and the risk output layer, and a quantitative dynamic risk assessment value with confidence interval is output. The risk factor input layer preprocesses the current risk factor data; the weight parsing layer obtains the basic weights of the current risk factor data based on the historical risk factor data and the judgment matrix; the fusion operation layer obtains the multi-factor joint fuzzy measure of the current risk factor data based on the fuzzy integral fusion model; and the risk output layer obtains and outputs the quantitative dynamic risk assessment value based on the evolutionary features and laws and the multi-factor joint fuzzy measure.

[0020] Specifically, obtaining the correlation between risk factors and accident probability is to clarify the degree of influence of different risk factors on accident occurrence, thereby distinguishing between core and secondary risk factors; extracting the spatial distribution, temporal variation, and dynamic change characteristics of historical risk factor data is to understand the changing patterns of risk factors in different road sections, time periods, and different stages of fog; identifying the combined effects of historical risk factor data is to clarify the synergistic or antagonistic effects between multiple factors. These are the core foundations for analyzing risk evolution; obtaining the evolutionary characteristics and patterns of risk based on the above characteristics and patterns is to ensure that the subsequently constructed risk fusion assessment framework conforms to the actual change logic of risks under foggy weather conditions on highways, making the framework dynamic. The rationality of dynamic and nonlinear assessments; obtaining the basic weights of current risk factor data based on historical risk factor data and judgment matrices is to objectively quantify the importance of each current risk factor by combining historical data, avoiding the bias of subjective assignment; obtaining the multi-factor joint fuzzy measure of current risk factor data based on the fuzzy integral fusion model is to accurately quantify the coupling effect between multiple risk factors and make up for the deficiency of traditional linear assessment in not being able to reflect the interaction of factors; and conducting risk assessment based on evolutionary characteristics and laws and multi-factor joint fuzzy measure is to ensure that the assessment results not only conform to the actual evolution trend of risk under foggy weather, but also reflect the actual impact of multi-factor coupling, and finally output an accurate quantitative dynamic risk assessment value.

[0021] Specifically, the historical risk factor dataset is preprocessed by first removing duplicate data and outliers exceeding reasonable physical limits, then using interpolation or historical averages of the same road segment and fog level to fill in missing data, and finally converting historical risk factor data from different collection formats into a unified numerical format and classifying them by time and road segment. The current risk factor data is preprocessed by first verifying the status of the collection equipment and the integrity of data transmission to remove invalid data, then using historical data from the same time period and road segment in the fog weather risk analysis database to fill in the current missing data, and finally converting the current data into a format consistent with the historical risk factor dataset to ensure compatibility for subsequent calculations.

[0022] Understandably, by extracting spatial distribution, temporal changes, and dynamic characteristics of fog stages from historical risk factor data, and matching the risk calculation logic of fog formation, development, and dissipation stages through a phased calculation submodule, the risk modeling is dynamically adjusted in stages. The weight analysis layer dynamically corrects the judgment matrix and adjusts factor weights by combining correlation and evolution patterns, while the fusion calculation layer dynamically quantifies the interaction relationship between core and secondary risk factors based on synergistic / antagonistic patterns. The risk output layer dynamically corrects the assessment value and outputs it by incorporating current risk factor data in real time for iterative optimization, and combines it with confidence intervals. This overcomes the limitations of traditional static modeling in adapting to the dynamic changes of fog risk, and significantly improves the real-time performance, accuracy, and scenario adaptability of risk assessment.

[0023] In some embodiments of this application, obtaining the correlation between risk factors and the probability of an accident includes: The historical risk factor dataset includes several historical risk factor data, including: atmospheric visibility, road surface friction coefficient, and traffic flow density. The first atmospheric visibility dataset is obtained by extracting the historical risk factor data corresponding to the maximum atmospheric visibility from the historical risk factor dataset, and the second atmospheric visibility dataset is obtained by extracting the historical risk factor data corresponding to the minimum atmospheric visibility from the historical risk factor dataset. The atmospheric visibility variation range is obtained by subtracting the atmospheric visibility minimum value from the maximum atmospheric visibility value. The atmospheric visibility accident probability variation range is obtained by subtracting the accident probability from the accident probability in the first atmospheric visibility dataset. The atmospheric visibility accident probability variation range is obtained by dividing the atmospheric visibility variation range by the atmospheric visibility accident probability variation range. The first road surface friction coefficient dataset is obtained by extracting the historical risk factor data corresponding to the maximum value of the road surface friction coefficient from the historical risk factor dataset, and the second road surface friction coefficient dataset is obtained by extracting the historical risk factor data corresponding to the minimum value of the road surface friction coefficient from the historical risk factor dataset. The change range of the road friction coefficient is obtained by subtracting the minimum value of the road friction coefficient from the maximum value of the road friction coefficient; the change range of the road friction coefficient accident probability is obtained by subtracting the accident probability of the second road friction coefficient dataset from the accident probability of the first road friction coefficient dataset; and the road friction coefficient correlation ratio is obtained by dividing the change range of the road friction coefficient by the change range of the road friction coefficient accident probability. The first traffic flow density dataset is obtained by extracting the historical risk factor data corresponding to the maximum traffic flow density from the historical risk factor dataset, and the second traffic flow density dataset is obtained by extracting the historical risk factor data corresponding to the minimum traffic flow density from the historical risk factor dataset. The traffic flow density change range is obtained by subtracting the minimum traffic flow density from the maximum traffic flow density. The traffic flow density accident probability change range is obtained by subtracting the accident probability from the accident probability in the first traffic flow density dataset. The traffic flow density correlation ratio is obtained by dividing the traffic flow density change range by the traffic flow density accident probability change range.

[0024] Specifically, the correlation ratio threshold is a critical value used to determine whether the three risk factors—atmospheric visibility, road surface friction coefficient, and traffic flow density—are strongly or weakly correlated with the probability of accident occurrence. It is an important basis for classifying core risk factors and secondary risk factors. When obtaining the correlation ratio threshold, the historical correlation ratios of each risk factor are first extracted from the fog weather risk analysis database, and extreme outliers are removed. Then, the 75th percentile of the effective correlation ratio of atmospheric visibility is taken as its correlation ratio threshold. For the correlation ratio of road surface friction coefficient, a curve showing the relationship between the correlation ratio and the frequency of accidents is plotted, and the abrupt change in slope is selected as its threshold. For the correlation ratio of traffic flow density, the average value is taken after combining expert scoring and weighting as its threshold.

[0025] Specifically, a correlation ratio threshold of 15 was set. From a historical risk factor dataset containing traffic operation and accident records under heavy fog conditions on highways over the past 5 years, atmospheric visibility was extracted, with a maximum value of 400 meters and a minimum value of 30 meters. The accident probability of the first atmospheric visibility dataset (historical risk factor data corresponding to atmospheric visibility of 400 meters) was 0.005, and the accident probability of the second atmospheric visibility dataset (historical risk factor data corresponding to atmospheric visibility of 30 meters) was 0.25. The atmospheric visibility variation was first calculated as 400 - 30 = 370 meters, and the atmospheric visibility accident probability variation was 0.005 - 0.25 = -0.245. The atmospheric visibility correlation ratio was then calculated as 370 ÷ 0.245 ≈ 1510.2. Subsequently, the road surface friction coefficient was extracted, with a maximum value of 0.85 and a minimum value of 0.15. The accident probability of the first road surface friction coefficient dataset (historical risk factor data corresponding to road surface friction coefficient of 0.85) was 0.01, and the second road surface friction coefficient dataset... The accident probability of the friction coefficient dataset (historical risk factor data corresponding to a road friction coefficient of 0.15) is 0.3. The calculated change range of the road friction coefficient is 0.85 - 0.15 = 0.7, and the change range of the accident probability of the road friction coefficient is 0.01 - 0.3 = -0.29. The correlation ratio of the road friction coefficient is 0.7 ÷ 0.29 ≈ 2.41. Further, the maximum traffic flow density is extracted as 220 vehicles / km and the minimum as 20 vehicles / km. The accident probability of the first traffic flow density dataset (historical risk factor data corresponding to a traffic flow density of 220 vehicles / km) is 0.02, and the accident probability of the second traffic flow density dataset (historical risk factor data corresponding to a traffic flow density of 20 vehicles / km) is 0.08. The calculated change range of the traffic flow density is 220 - 20 = 200 vehicles / km, and the change range of the accident probability of the traffic flow density is 0.02 - 0.08 = -0.06. The correlation ratio of the traffic flow density is 200 ÷ 0.06 ≈ 3333.33.

[0026] Understandably, by quantifying the ratio of the change in each factor to the change in the probability of the accident, a precise quantitative basis is provided for subsequently determining the strength of the correlation between factors and accidents. This operation achieves a preliminary dynamic quantification of the impact of risk factors, avoiding the drawbacks of subjectively judging the importance of factors in traditional assessments. It is a key prerequisite for subsequently dynamically classifying core and secondary risk factors and constructing a dynamic weighting system.

[0027] In some embodiments of this application, when obtaining the correlation between risk factors and the probability of an accident, the method further includes: When the atmospheric visibility correlation ratio is greater than the correlation ratio threshold, the correlation between atmospheric visibility and the probability of an accident is determined to be a strong correlation, and atmospheric visibility is marked as a core risk factor; when the atmospheric visibility correlation ratio is less than or equal to the correlation ratio threshold, the correlation between atmospheric visibility and the probability of an accident is determined to be a weak correlation, and atmospheric visibility is marked as a secondary risk factor. When the correlation ratio of the road surface friction coefficient is greater than the correlation ratio threshold, the correlation between the road surface friction coefficient and the probability of accident occurrence is determined to be a strong correlation, and the road surface friction coefficient is marked as a core risk factor; when the correlation ratio of the road surface friction coefficient is less than or equal to the correlation ratio threshold, the correlation between the road surface friction coefficient and the probability of accident occurrence is determined to be a weak correlation, and the road surface friction coefficient is marked as a secondary risk factor. When the traffic flow density correlation ratio is greater than the correlation ratio threshold, the correlation between the traffic flow density and the probability of an accident is determined to be strong, and the traffic flow density is marked as a core risk factor; when the traffic flow density correlation ratio is less than or equal to the correlation ratio threshold, the correlation between the traffic flow density and the probability of an accident is determined to be weak, and the traffic flow density is marked as a secondary risk factor.

[0028] Specifically, a correlation ratio threshold of 15 was set. Based on a historical risk factor dataset containing traffic operation and accident records under heavy fog conditions on highways over the past 5 years, the correlation ratio for atmospheric visibility was calculated to be approximately 1510.2, which is greater than the correlation ratio threshold of 15. Therefore, atmospheric visibility was determined to have a strong correlation with the probability of accidents, and atmospheric visibility was marked as a core risk factor. Next, the correlation ratio for road surface friction coefficient was calculated to be approximately 2.41, which is less than or equal to the correlation ratio threshold of 15. Therefore, road surface friction coefficient was determined to have a weak correlation with the probability of accidents, and road surface friction coefficient was marked as a secondary risk factor. Finally, the correlation ratio for traffic flow density was calculated to be approximately 3333.33, which is greater than the correlation ratio threshold of 15. Therefore, traffic flow density was determined to have a strong correlation with the probability of accidents, and traffic flow density was marked as a core risk factor.

[0029] Understandably, determining the correlation strength between each risk factor and the probability of accident occurrence based on the correlation ratio threshold, and labeling core and secondary risk factors, clarifies the hierarchical positioning of different factors in risk assessment. This classification method provides a clear basis for the differentiated treatment of core and secondary factors in subsequent dynamic modeling (such as weight adjustment and quantification of interaction relationships), enabling risk assessment to focus on the dynamic changes of core risk factors, improving the pertinence and efficiency of dynamic modeling, and avoiding assessment bias caused by treating all factors equally.

[0030] In some embodiments of this application, extracting the spatial distribution characteristics, temporal variation characteristics, and dynamic variation trends of historical risk factor data, and identifying the combined interaction patterns among the historical risk factor data, includes: The historical risk factor dataset is divided into subsets of different regions according to spatial road segments. The mean values ​​of each subset are calculated at different time points, and the changes in the mean values ​​reflect the spatial distribution characteristics. The numerical fluctuation values ​​of the same subset are calculated over a continuous period of time, and the numerical fluctuation values ​​reflect the temporal change characteristics. The numerical change curves of the historical risk factor data during the formation, development, and dissipation stages of fog are fitted using time series analysis methods, and the dynamic change trends are extracted by the slope of the numerical change curves. The core risk factor is combined with the secondary risk factor to obtain multiple risk factor combinations, and each risk factor combination contains at least two individual risk factors with different relationships. The historical risk factor data of the accidents that occurred are extracted from the historical risk factor dataset to obtain the accident dataset; The risk factor combination is extracted from the accident dataset to obtain the accident combination dataset, and the risk factor combination in the accident combination dataset is the accident risk factor combination; the risk factor combination is extracted from the historical risk factor dataset to obtain the overall combination dataset; the proportion of the accident risk factor combination in the overall combination dataset is the probability of occurrence of the risk factor combination accident; The individual risk factor is extracted from the accident dataset to obtain a single risk factor dataset, and the individual risk factor in the accident combination dataset is a single accident risk factor; the individual risk factor is extracted from the historical risk factor dataset to obtain an overall single risk factor dataset; the proportion of the individual accident risk factor in the overall single risk factor dataset is the probability of occurrence of the single risk factor accident.

[0031] Specifically, the historical risk factor dataset is defined as a collection of traffic data from the K1-K3 section of the highway during the past 5 years under foggy weather conditions. This dataset includes hourly atmospheric visibility, road surface friction coefficient, traffic flow density, and accident records for each road segment. The core risk factors are atmospheric visibility and traffic flow density, while the secondary risk factor is the road surface friction coefficient. First, the dataset is divided into three sub-datasets based on spatial road segments: K1 (containing 2000 data points), K2 (containing 2200 data points), and K3 (containing 1800 data points). The average values ​​for each sub-dataset are calculated at three time points: the 1st, 10th, and 20th of each month. For example, K... The average atmospheric visibility values ​​in Zone 1 were 50 meters, 40 meters, and 30 meters, respectively; in Zone 2, they were 80 meters, 75 meters, and 70 meters; and in Zone 3, they were 60 meters, 65 meters, and 60 meters. These changes in averages reflect a spatial distribution characteristic of continuously decreasing risk factor values ​​in Zone 1, a slow decreasing characteristic in Zone 2, and small fluctuations in Zone 3. Further calculations of the numerical fluctuations of the same subset over continuous time periods were performed. For example, the traffic flow density fluctuation during foggy periods was 30 vehicles / km in Zone 1, 15 vehicles / km in Zone 2, and 20 vehicles / km in Zone 3. These fluctuations show that the temporal variation characteristics of Zone 1 are more significant. Finally, time series analysis methods were used... Historical risk factor data were fitted to the numerical change curves of atmospheric visibility during the fog formation (0-6h), development (6-12h), and dissipation (12-24h) stages. The slope of the curve was -8 m / h during the formation stage, -1 m / h during the development stage, and 5 m / h during the dissipation stage. The slope was used to extract the dynamic trend of atmospheric visibility decreasing rapidly during the fog formation stage, decreasing slowly during the development stage, and increasing rapidly during the dissipation stage. Subsequently, the core risk factors were combined with secondary risk factors to obtain risk factor combination 1 (atmospheric visibility + road friction coefficient), combination 2 (traffic flow density + road friction coefficient), and combination 3 (atmospheric visibility + traffic flow density). Three combinations are identified. Then, 100 accident data points are extracted from the historical risk factor dataset to form the accident dataset. From the accident dataset, the aforementioned risk factor combinations are extracted to obtain the accident combination dataset (combination 1 occurs 40 times, combination 2 occurs 30 times, and combination 3 occurs 20 times). All risk factor combinations are extracted from the historical risk factor dataset to obtain the overall combination dataset (combination 1 occurs 1000 times, combination 2 occurs 900 times, and combination 3 occurs 800 times). The calculated probabilities of accidents occurring for each risk factor combination are: Combination 1: 40 / 1000 = 4%, Combination 2: 30 / 900 ≈ 3.33%, and Combination 3: 20 / 800 = 2%.5%; Finally, individual risk factors are extracted from the accident dataset to obtain a single risk factor dataset (atmospheric visibility occurring 100 times, road friction coefficient occurring 100 times, and traffic flow density occurring 100 times). Individual risk factors are also extracted from the historical risk factor dataset to obtain a total single risk factor dataset (atmospheric visibility occurring 5000 times, road friction coefficient occurring 5000 times, and traffic flow density occurring 5000 times). The probability of an accident occurring for each individual risk factor is calculated to be 100 / 5000 = 2%. This completes all the operations before extracting features and identifying combined action patterns in claim 4.

[0032] Understandably, the system successfully extracted the spatial distribution, temporal variations, and dynamic characteristics of historical risk factor data during foggy periods, as well as prepared relevant data on risk factor combinations. This allowed for the multi-dimensional mining of risk dynamics from both spatiotemporal perspectives and fog evolution stages. These feature extraction results provided data support for the subsequent staged computational submodules to match the computational logic for different fog stages. They also laid the foundation for analyzing the combination patterns between factors, enabling dynamic modeling to align with the spatiotemporal dynamic changes of risk under foggy weather conditions, rather than being limited to static analysis within a single dimension.

[0033] In some embodiments of this application, when extracting the spatial distribution characteristics, temporal variation characteristics, and dynamic variation trends of historical risk factor data, and identifying the combined interaction patterns among the historical risk factor data, the method further includes: When the probability of occurrence of the combined risk factor event is greater than or equal to the sum of the probabilities of occurrence of each individual risk factor event, the combined action mode of the current risk factor combination is a synergistic action mode. When the probability of an event occurring in the combination of risk factors is less than the sum of the probabilities of each individual risk factor event, the combined action mode of the current risk factor combination is an antagonistic action mode.

[0034] Specifically, the probability of an accident for risk factor combination 1 (atmospheric visibility + road friction coefficient) is 4%, the probability of an accident for combination 2 (traffic flow density + road friction coefficient) is approximately 3.33%, the probability of an accident for combination 3 (atmospheric visibility + traffic flow density) is 2.5%, and the probability of an accident for each individual risk factor (atmospheric visibility, road friction coefficient, traffic flow density) is 2%. First, the sum of the accident probabilities of each individual risk factor in combination 1 is calculated as 2% + 2% = 4%. Since the accident probability of 4% for combination 1 is greater than or equal to this sum, the combined effect mode of combination 1 is determined to be a synergistic effect mode. Next, the sum of the accident probabilities of each individual risk factor in combination 2 is calculated as 2% + 2% = 4%. Since the accident probability of 3 for combination 2 is 3.33%, less than this sum, the combined effect mode of combination 2 is determined to be an antagonistic effect mode. Finally, the sum of the accident probabilities of each individual risk factor in combination 3 is calculated as 2% + 2% = 4%. Since the accident probability of 2.5% for combination 3 is less than this sum, the combined effect mode of combination 3 is determined to be an antagonistic effect mode.

[0035] Understandably, by comparing the sum of the probabilities of combined accidents and individual factor accidents, the synergistic or antagonistic effects among risk factors can be identified, clarifying the dynamic interaction types between factors. This identification result is the core premise for the fusion computing layer to dynamically quantify factor relationships based on interaction patterns. It breaks through the limitations of traditional assessments that only consider the effects of a single factor and ignore the interaction effects between factors, enabling dynamic modeling to more realistically reflect the actual risk situation of multiple risk factors acting together under foggy weather conditions.

[0036] In some embodiments of this application, when obtaining the evolutionary characteristics and patterns of risk based on the spatial distribution characteristics, the temporal variation characteristics, the dynamic change trend, and the combined action mode, the following methods are included: Based on the spatial distribution characteristics, the correspondence between the mean values ​​of the subsets in different regions and the frequency of accidents is statistically analyzed. When the change in the mean value is greater than a first threshold, the risk in the current region is more affected by the synergistic effect mode than by the antagonistic effect mode. When the change in the mean value is less than or equal to the first threshold, the risk in the current region is more affected by the antagonistic effect mode than by the synergistic effect mode. Based on the time change characteristics, the correspondence between the numerical fluctuation values ​​of the same subset of data and the time period of the accident is statistically analyzed. When the numerical fluctuation value is greater than the fluctuation value threshold, the risk in the current period fluctuates rapidly with the change in the value of the core risk factor; when the numerical fluctuation value is less than or equal to the fluctuation value threshold, the risk in the current period remains stable. Based on the dynamic change trend and the slope of the numerical change curve, the risk change trend at different stages of fog is summarized. When the fog is in the formation stage, the risk increases with the increase of the slope; when the fog is in the development stage, the risk stabilizes as the slope approaches zero; and when the fog is in the dissipation stage, the risk decreases as the slope decreases. Based on the correlation between the numerical changes of the risk factors and the combined action patterns, the evolution law of risk with historical risk factor data is summarized. When the value of the core risk factor is in the extreme range and the core risk factor and the secondary risk factor form a synergistic action pattern, the risk changes exponentially; when the value of the core risk factor is in the normal range and the core risk factor and the secondary risk factor form the antagonistic action pattern, the risk changes linearly.

[0037] Specifically, the first amplitude threshold is used to determine the degree to which the mean change of numerical values ​​in different regional subsets indicates that the current regional risk is more significantly affected by synergistic or antagonistic patterns. The volatility threshold is used to determine whether the risk in the current period fluctuates rapidly with the core risk factor or remains stable when the numerical volatility of the same subset exceeds or falls below this value. To obtain the first amplitude threshold, the mean change of numerical values ​​in all regional subsets is extracted from the fog weather risk analysis database. After removing extreme outliers, the remaining data is sorted in ascending order, and the 80th percentile of the sorted data is taken as the first amplitude threshold. To obtain the volatility threshold, the numerical volatility of each subset in the historical risk factor dataset is calculated. Combined with the risk change records during the accident period, the numerical volatility corresponding to the risk transitioning from a stable state to a rapidly fluctuating state is selected as the volatility threshold.

[0038] Specifically, based on the subsets of K1, K2, and K3 regions, a first amplitude threshold of 15 meters (the critical value for the mean change in atmospheric visibility) and a fluctuation threshold of 20 vehicles / km (the critical value for the fluctuation in traffic flow density) were set. The extreme range for the core risk factor atmospheric visibility was 0-50 meters, and the normal range was 50-500 meters. The extreme range for traffic flow density was 200-250 vehicles / km, and the normal range was 20-200 vehicles / km. First, based on spatial distribution characteristics, the mean change in the K1 region subset was statistically determined to be 20 meters, 10 meters for K2, and 8 meters for K3. The correlation between the frequency of accidents in each region (80 times in region K1, 30 times in region K2, and 25 times in region K3) was established. Since the mean value change in region K1 (20 meters) exceeded the first threshold of 15 meters, it was determined that the risk in region K1 was more influenced by the synergistic effect mode than the antagonistic effect mode. The mean value changes in regions K2 and K3 were less than or equal to the first threshold, indicating that the risk in these two regions was more influenced by the antagonistic effect mode than the synergistic effect mode. Then, based on time-varying characteristics, the statistical values ​​for the subset data were 30 vehicles / km in region K1, 15 vehicles / km in region K2, and 25 vehicles / km in region K3. The correlation between the occurrence of accidents in different time periods (60 accidents in K1 area, 10 in K2 area, and 20 in K3 area) and the numerical fluctuation values ​​in K1 and K3 areas exceeding the fluctuation threshold of 20 vehicles / km indicates that the risk in these two areas is rapidly fluctuating with the changes in the core risk factors. The numerical fluctuation value in K2 area is less than or equal to the fluctuation threshold, indicating that its risk remains stable. Furthermore, based on the dynamic trend, and considering the slope of the atmospheric visibility curve during the fog formation stage (0-6h) (-8 m / h), the development stage (6-12h) (-1 m / h), and the dissipation stage (... The slope of the fog (12-24h) is 5 m / h. It is concluded that the risk increases with the increase of the slope during the fog formation stage, remains stable as the slope approaches zero during the development stage, and decreases as the slope decreases during the dissipation stage. Finally, based on the correlation between the numerical changes of risk factors and the combined action mode, when the core risk factor, atmospheric visibility, is in the extreme range of 0-50 meters and forms a synergistic effect with the secondary risk factor, road friction coefficient, the risk is judged to change exponentially. When atmospheric visibility is in the normal range of 50-500 meters and forms an antagonistic effect with the road friction coefficient, the risk is judged to change linearly.

[0039] Understandably, based on the extracted dynamic features and combination patterns, the evolutionary characteristics and patterns of risk were obtained, revealing the risk change patterns in different regions, time periods, and different fog stages, as well as the impact of factor numerical ranges and interaction patterns on risk change trends. These evolutionary patterns provide a theoretical basis for the phased calculation submodule to match the risk calculation logic of fog formation, development, and dissipation stages, giving the phased adjustment of dynamic modeling a scientific and systematic support, and ensuring that the assessment model can dynamically adapt to the evolution of fog stages.

[0040] In some embodiments of this application, when constructing a risk fusion assessment framework based on the aforementioned evolutionary characteristics and patterns, the following are included: A regionalized data processing submodule is set up in the risk fusion assessment framework, and risk factor data for different road sections are processed differently. A phased calculation submodule is set up in the risk fusion assessment framework, and the risk calculation logic is matched to the formation, development and dissipation stages of fog. A coupling effect quantification submodule is set up in the framework to quantify the interaction relationship between the core risk factor and the secondary risk factor under the synergistic and antagonistic modes.

[0041] Specifically, the risk factor data for different road segments is processed differently based on the evolutionary characteristics of risk across different road segments. Different data collection frequencies and weightings are set for different road segments such as K1, K2, and K3. For example, for K1 road segments where the mean value change exceeds the first threshold and is more affected by synergistic effects, the core risk factor data collection frequency is increased to once every 5 minutes and the weighting is increased to 70%. For K2 and K3 road segments, the collection frequency is maintained at once every 15 minutes, and the core risk factor weighting is set at 50%. Simultaneously, the calculation dimensions of risk factors are adjusted based on the road characteristics of each segment (such as curves and slopes). Matching the risk calculation logic to the formation, development, and dissipation stages of fog involves establishing a high-weight calculation logic (using an exponential calculation formula) for the core risk factor numerical change rate during the formation stage, and a numerical stability calculation logic for the development stage. The system employs a priority-based weighted calculation logic (using a linear stationary operation formula) and a decreasing weighted calculation logic (using a decreasing exponential operation formula) for the numerical recovery rate during the dissipation phase. It automatically matches the calculation logic and iteration parameters corresponding to the current fog phase through a phased operation submodule. To quantify the interaction between core and secondary risk factors under synergistic and antagonistic modes, a coupling coefficient is introduced as a quantitative indicator. Under synergistic mode, the coupling coefficient is set to 1.5, and the interaction gain effect is quantified by multiplying the product of the core and secondary risk factors by this coupling coefficient. Under antagonistic mode, the coupling coefficient is set to 0.6, and the interaction offsetting effect is quantified by multiplying the sum of the core and secondary risk factors by this coupling coefficient. Simultaneously, the coupling coefficient is dynamically adjusted based on the difference between the probability of a combined risk factor event and the probability of an individual risk factor event, thereby achieving precise quantification of factor interaction relationships.

[0042] Understandably, a risk fusion assessment framework was established, comprising sub-modules for regionalized data processing, phased computation, and coupling effect quantification. This framework enables differentiated processing of data from different road segments, dynamic matching of computational logic across different stages of fog, and quantification of interactions between factors. This framework serves as the core carrier for dynamic modeling. The regionalized sub-module allows the assessment to adapt to the risk characteristics of different road segments, the phased sub-module enables dynamic modeling of the entire fog lifecycle, and the coupling effect sub-module provides tools for accurately quantifying factor interactions, comprehensively improving the dynamic scenario adaptability of the assessment method.

[0043] In some embodiments of this application, when the weight parsing layer obtains the basic weights of the current risk factor data based on the historical risk factor data and the judgment matrix, it includes: An initial judgment matrix is ​​constructed based on the risk factor types of the current risk factor data. The rows and columns of the initial judgment matrix correspond to the risk factor types of the current risk factor data, and the matrix elements are assigned values ​​to represent the relative importance of different current risk factor data. The correlation results corresponding to the historical risk factor data and the evolution law of risk with the historical risk factor data are retrieved from the fog weather risk analysis database. The correlation results and the evolution law are used together as correction coefficients to adjust the matrix elements of the initial judgment matrix to obtain the corrected judgment matrix. The consistency ratio of the corrected judgment matrix is ​​calculated. When the consistency ratio is greater than the consistency threshold, the matrix elements of the initial judgment matrix are readjusted until the consistency ratio is less than or equal to the consistency threshold. When the consistency ratio is less than or equal to the consistency threshold, the current risk factor data is normalized. For negative factor data such as atmospheric visibility, the reciprocal standardization method is used. For positive factor data such as road friction coefficient and traffic flow density, the extreme value method is used to map to a unified numerical range. The normalized current risk factor data is substituted into the corrected judgment matrix. The maximum eigenvalue and corresponding eigenvector of the corrected judgment matrix are calculated using the eigenvalue method. The eigenvector is normalized to obtain the basic weight of the current risk factor data. The weight adjustment rules for different stages of fog are matched through the staged operation submodule, and the adjusted basic weight is passed to the fusion operation layer.

[0044] Specifically, based on the current risk factor data, the risk factor types are atmospheric visibility, road friction coefficient, and traffic flow density. A 3×3 initial judgment matrix is ​​constructed using the 1-9 scaling method. The rows and columns of the matrix correspond to these three types of factors. Values ​​are assigned to each factor according to its empirical importance. For example, atmospheric visibility is assigned a value of 1, road friction coefficient a value of 3, and traffic flow density a value of 2; road friction coefficient is assigned a value of 1 / 3 to atmospheric visibility, a value of 1 to itself, and a value of 1 / 2 to traffic flow density; traffic flow density is assigned a value of 1 / 2 to atmospheric visibility, a value of 2 to road friction coefficient, and a value of 1 to itself. This forms the initial judgment matrix. The correlation results are then... The strong correlation correction coefficient of atmospheric visibility and traffic flow density is 1.2, and the weak correlation correction coefficient of road friction coefficient is 0.8. Combined with the risk impact correction coefficient of 1.3 in the fog formation stage of the evolution law, the comprehensive correction coefficients of each factor are calculated (atmospheric visibility: 1.2×1.3=1.56, road friction coefficient: 0.8×1.3=1.04, traffic flow density: 1.2×1.3=1.56). Then, the comprehensive correction coefficients of each factor are multiplied by the matrix elements of the corresponding row in the initial judgment matrix to complete the adjustment of the initial judgment matrix and obtain the corrected judgment matrix. When calculating the consistency ratio of the corrected judgment matrix, First, the largest eigenvalue λmax of the corrected matrix is ​​obtained using the eigenvalue method. Then, the consistency index CI = (λmax - n) / (n - 1) (where n is the matrix order, and here n = 3) is calculated. Next, the random consistency index RI is queried (RI = 0.58 when n = 3). Finally, the consistency ratio is calculated using CR = CI / RI. If CR < 0.1, the matrix is ​​considered to meet the consistency requirements. When normalizing the current risk factor data, for negative factor data such as atmospheric visibility, the reciprocal of the value is taken first (e.g., if the current atmospheric visibility is 50 meters, the reciprocal is 1 / 50 = 0.02). Then, the (z-min) is used to calculate the consistency ratio. The standardized formula for / (max-min) (where z is the reciprocal value, and max and min are historical extreme values ​​after taking the reciprocal) is standardized. For positive factor data such as road friction coefficient and traffic flow density, the extreme value method is used to map to a unified numerical range of [0,1]. For example, if the current road friction coefficient is 0.5, and its historical extreme values ​​are 0.1 and 0.9, then the mapped value is (0.5-0.1) / (0.9-0.1)=0.5. If the current traffic flow density is 100 vehicles / km, and its historical extreme values ​​are 20 vehicles / km and 220 vehicles / km, then the mapped value is (100-20) / (220-20)=0.4. When calculating the maximum eigenvalue and corresponding eigenvector of the corrected judgment matrix using the eigenvalue method, first construct the determinant |A-λI|=0 (A is the corrected judgment matrix, λ is the eigenvalue, and I is the identity matrix). Solve this determinant to obtain the eigenvalues, where the largest eigenvalue is the maximum eigenvalue λmax. Then substitute λmax into the homogeneous linear equation system (A-λmaxI)X=0 (X is the eigenvector), and solve this system to obtain the eigenvector corresponding to the maximum eigenvalue. When normalizing the eigenvector, divide each element of the eigenvector by the sum of all elements. The result is the basic weight of the current risk factor data (e.g., if the eigenvector is [0.6, 0.2, 0.7], the sum of its elements is 1.5). After normalization, the base weights are [0.4, 0.133, 0.467]. Finally, the weight adjustment rules for different stages of fog are matched through the staged calculation submodule. During the fog formation stage, the base weights of the core risk factors (atmospheric visibility and traffic flow density) are increased by 20% (atmospheric visibility 0.4 × 1.2 = 0.48, traffic flow density 0.467 × 1.2 = 0.56), while the base weight of the secondary risk factor (road friction coefficient) is decreased by 10% (0.133 × 0.9 = 0.12). During the fog development stage, the base weights remain unchanged. During the fog dissipation stage, the base weights of the core risk factors are decreased by 15%, while the base weights of the secondary risk factors are increased by 10%, thus completing the weight adjustment for different stages.

[0045] Specifically, the consistency ratio is a quantitative indicator that measures whether the judgment logic of each element in the judgment matrix is ​​consistent and whether there are contradictions. It is the ratio of the consistency index to the random consistency index, and its value directly reflects the degree of consistency of the judgment matrix. The consistency threshold is the critical value for determining whether the judgment matrix has acceptable consistency. It is the upper limit standard of the consistency ratio of the judgment matrix without adjustment. When obtaining the consistency ratio, the largest eigenvalue of the judgment matrix is ​​first obtained by using the eigenvalue method. Then, the consistency index is calculated according to the formula CI=(λmax-n) / (n-1) (where λmax is the largest eigenvalue and n is the matrix order). Next, the random consistency index RI for the corresponding matrix order is looked up. Finally, the consistency ratio is calculated by CR=CI / RI. When obtaining the consistency threshold, no complex calculation is required. The general standard of the analytic hierarchy process, which has been verified by a large amount of mathematical statistics and practice, is directly adopted. For judgment matrices of order 3-9, 0.1 is determined as the consistency threshold.

[0046] Specifically, the risk factor types for the current risk factor data are set as atmospheric visibility (A), road surface friction coefficient (B), and traffic flow density (C). First, a 3×3 initial judgment matrix is ​​constructed using the 1-9 scaling method. The rows and columns of the matrix correspond to the three types of factors A, B, and C. Based on empirical importance, values ​​are assigned as follows: A to A = 1, A to B = 5, A to C = 3; B to A = 1 / 5, B to B = 1, B to C = 1 / 3; C to A = 1 / 3, C to B = 3, C to C = 1. This completes the construction of the initial judgment matrix. Next, the correlation results (A and C are strongly correlated, with a correction coefficient of 1.4; B is weakly correlated, with a correction coefficient of 0.7) and risk evolution patterns (for fog) are retrieved from the fog weather risk analysis database. The formation stage correction coefficient is 1.2). Calculate the comprehensive correction coefficients for each factor: A = 1.4 × 1.2 = 1.68, B = 0.7 × 1.2 = 0.84, C = 1.4 × 1.2 = 1.68. Then multiply the comprehensive correction coefficients by the elements of the corresponding rows of the initial judgment matrix to obtain the corrected judgment matrix: the first row is [1.68, 8.4, 5.04], the second row is [0.168, 0.84, 0.28], and the third row is [0.56, 5.04, 1.68]. Subsequently, calculate the consistency ratio of the corrected judgment matrix. First, obtain the maximum eigenvalue of the corrected matrix λmax ≈ 3.05 using the eigenvalue method. Combined with the matrix order n = 3, calculate the consistency ratio using the formula CI = (λmax - ... The consistency index CI is calculated as 0.025 using (n) / (n-1). The random consistency index RI of the 3rd order matrix is ​​found to be 0.58. The consistency ratio CR is then calculated as CR≈0.043 using CR=CI / RI. A consistency threshold of 0.1 is set. Since 0.043<0.1, the matrix meets the consistency requirements, and no adjustment to the initial matrix elements is needed. Next, the current risk factor data is normalized. For negative factors such as atmospheric visibility, the current value is 40 meters. The reciprocal is taken to obtain 1 / 40=0.025. Then, based on the historical extreme values ​​after taking the reciprocal (max=0.05, min=0.01), the standardized value is calculated using the formula (z-min) / (max-min). The value is 0.375. For positive factors such as road surface friction coefficient and traffic flow density, the current road surface friction coefficient is 0.6. Combining its historical extreme values ​​(max=0.9, min=0.1), the extreme value method is used to map it to the interval [0,1] to get (0.6-0.1) / (0.9-0.1)=0.625. The current traffic flow density is 150 vehicles / km. Combining its historical extreme values ​​(max=220, min=20), the extreme value method is used to map it to get (150-20) / (220-20)=0.65. Then, the normalized current risk factor data is substituted into the corrected judgment matrix, and the determinant |A-λI|=0 is solved by the eigenvalue method to obtain the eigenvalue. The largest λmax≈3 is taken.05 is taken as the largest eigenvalue. Substituting it into the homogeneous linear equation system (A-λmaxI)X=0, the corresponding eigenvector is obtained as [0.72, 0.15, 0.81]. This eigenvector is then normalized. First, the element-wise sum is calculated as 0.72+0.15+0.81=1.68. Then, each element is divided by the sum to obtain the basic weights of the current risk factor data: A≈0.429, B≈0.089, C≈0.482. Finally, the weight adjustment rules for different stages of the fog are matched through the staged operation submodule. During the fog formation stage, the basic weights of core risk factors A and C are increased by 20% (A≈0.515, C≈0.578), and the basic weight of secondary risk factor B is decreased by 10% (B≈0.080). During the fog development stage, the basic weights remain unchanged. During the fog dissipation stage, the weights of core risk factors are decreased by 15%, and the weights of secondary risk factors are increased by 10%. The adjusted basic weights are then passed to the fusion operation layer.

[0047] Understandably, the judgment matrix is ​​dynamically revised at the weight analysis layer by combining the correlation results with the risk evolution pattern. Basic weights are obtained through consistency checks and normalization, and then adjusted according to the fog stage. This process overcomes the shortcomings of fixed weights in traditional assessments, achieving dynamic revision and staged adjustment of weights. This allows the weights of core risk factors to dynamically adapt to changes in the fog stage and the strength of factor correlations, making the dynamically modeled weight system more closely reflect actual risk conditions and improving the accuracy of the assessment.

[0048] In some embodiments of this application, when the fusion operation layer obtains the multi-factor joint fuzzy measure of the current risk factor data based on a fuzzy integral fusion model, it includes: The coupling effect quantification submodule uses the adjusted base weight values ​​and the evolution pattern of risk with the historical risk factor data to determine the single-factor fuzzy measure value of each current risk factor data. The single-factor fuzzy measure value is positively correlated with the base weight values. The fluctuation range of the single-factor fuzzy measure value is calculated. When the fluctuation range of the measure value is greater than the second fluctuation threshold, the single-factor fuzzy measure value is smoothed. When the fluctuation range of the measure value is less than or equal to the second fluctuation threshold, a multi-factor joint fuzzy measure is calculated based on the single-factor fuzzy measure value. When calculating the multi-factor joint fuzzy measure, the synthesis rules of fuzzy integrals are combined with the synergistic and antagonistic effects to perform a weighted summation on multiple single-factor fuzzy measure values. The weight coefficients of the weighted summation are adjusted based on the combined effect mode of the risk factor combination. When the combined effect mode is the synergistic effect mode, the corresponding weight coefficient is increased; when the combined effect mode is the antagonistic effect mode, the corresponding weight coefficient is decreased. After the calculation is completed, the multi-factor joint fuzzy measure is passed to the risk output layer.

[0049] Specifically, the second fluctuation threshold is a critical value used to determine whether the fluctuation range of a single-factor fuzzy measure value exceeds a reasonable range and whether smoothing processing of the single-factor fuzzy measure value is necessary. It is an important indicator for measuring the stability of a single-factor fuzzy measure value. When obtaining the second fluctuation threshold, all fluctuation range data of historical single-factor fuzzy measure values ​​are first extracted from the fog weather risk analysis database. After removing extreme outliers, the remaining valid data are sorted in ascending order, and the 90th percentile of the sorted data is taken as the second fluctuation threshold. Alternatively, the fluctuation range value that minimizes the calculation error of the multi-factor joint fuzzy measure can be selected as the second fluctuation threshold through multiple simulation calculations, taking into account the computational accuracy requirements of the fuzzy integral fusion model.

[0050] Specifically, the second fluctuation threshold is set to 0.15, and the adjusted base weight values ​​are atmospheric visibility 0.515, road surface friction coefficient 0.080, and traffic flow density 0.578 (adjusted after the fog formation stage). Combining the evolution pattern of risk with historical risk factor data (the correction coefficient for the core risk factor measurement value during the fog formation stage is 1.2, and for the secondary risk factor it is 0.9), the single-factor fuzzy measurement value of each current risk factor data is first determined through the coupling effect quantification submodule: atmospheric visibility = 0.515 × 1.2 = 0. 618. Road surface friction coefficient = 0.080 × 0.9 = 0.072, traffic flow density = 0.578 × 1.2 = 0.694, and these three single-factor fuzzy measure values ​​are positively correlated with their corresponding basic weight values. Next, the fluctuation range of the single-factor fuzzy measure values ​​is calculated. The range of the three values ​​is 0.694 - 0.072 = 0.622 (standard deviation approximately 0.279). This fluctuation range is greater than the second fluctuation threshold of 0.15. Therefore, the moving average method is used to smooth the single-factor fuzzy measure values. The atmospheric visibility is 0.60, the road surface friction coefficient is 0.08, and the traffic flow density is 0.68. Then, when calculating the multi-factor joint fuzzy measure, the Choquet synthesis rule of the fuzzy integral is combined with the combined action modes determined in claim 5 (Combination 1: atmospheric visibility + road surface friction coefficient is a synergistic action mode; Combination 2: traffic flow density + road surface friction coefficient is an antagonistic action mode; Combination 3: atmospheric visibility + traffic flow density is an antagonistic action mode). The weighting coefficient for the synergistic action mode is increased to 1.2. The weight coefficients of the combination of antagonistic modes are reduced to 0.8, and then the weighted sum of multiple single-factor fuzzy measure values ​​is calculated: 0.60×1.2 (coefficient of combination 1) + 0.08×(1.2+0.8) (participating in combinations 1 and 2) + 0.68×(0.8+0.8) (participating in combinations 2 and 3) = 0.72 + 0.16 + 1.088 = 1.968. After normalization by the fuzzy integral formula, the multi-factor joint fuzzy measure value is 0.656. Finally, the multi-factor joint fuzzy measure is passed to the risk output layer.

[0051] Understandably, the fusion operation layer determines the single-factor fuzzy measure value based on the adjusted weights and evolutionary patterns, smooths unstable values ​​according to the second fluctuation threshold, and calculates the multi-factor joint fuzzy measure by adjusting the weight coefficients in conjunction with the synergistic / antagonistic mode. This operation achieves dynamic quantification of factor interaction relationships, improves the stability of the measure value through smoothing, and makes the fusion operation more closely reflect the actual interaction relationships between factors by adjusting the coefficients according to the interaction mode. This allows the dynamic modeling fusion stage to accurately reflect the comprehensive risk under multi-factor synergy or antagonism, avoiding the evaluation error caused by static fusion.

[0052] In some embodiments of this application, when the risk output layer obtains and outputs the quantitative dynamic risk assessment value based on the evolutionary features and patterns and the multi-factor joint fuzzy measure, it includes: Based on the evolution characteristics of risk with spatial road segments and time dimensions, as well as the evolution law of risk with the historical risk factor data, a corresponding nonlinear calculation method is selected, and the multi-factor joint fuzzy measure is substituted into the nonlinear calculation method for initial calculation to obtain a preliminary risk assessment value. The current risk factor data is incorporated into the nonlinear calculation method in real time for iterative optimization. The iterative parameters are adjusted by combining the dynamic changes in the formation, development and dissipation stages of the fog through the staged calculation submodule, and the deviation of the preliminary risk assessment value before and after iterative optimization is calculated. When the deviation value is greater than the deviation threshold, new current risk factor data is continuously incorporated and the iteration parameters are adjusted in conjunction with the spatial distribution characteristics until the deviation value is less than or equal to the deviation threshold; when the deviation value is less than or equal to the deviation threshold, the iteration is stopped and the confidence interval of the preliminary risk assessment value is calculated. When calculating the confidence interval, the standard deviation is calculated using the numerical distribution of the iteratively optimized risk assessment value. The weight of the standard deviation is adjusted in combination with the risk fluctuation characteristics of different regions in the spatial distribution characteristics. The upper and lower limits of the confidence interval are determined based on the adjusted standard deviation. The iteratively optimized risk assessment value is combined with the confidence interval to form the quantitative dynamic risk assessment value with confidence interval and output.

[0053] Specifically, the deviation threshold is a critical value used to determine whether the deviation of the initial risk assessment value before and after iterative optimization is within an acceptable range and whether it is necessary to continue to incorporate new current risk factor data for iteration. It is the core indicator for deciding whether to stop the risk assessment iteration process. When obtaining the deviation threshold, all deviation value data in the historical risk assessment iteration process are first extracted from the fog weather risk analysis database. After removing extreme outliers, the remaining valid data are sorted in ascending order. Combined with the accuracy requirements of risk assessment (such as the assessment error needing to be controlled within 5%), the 80th percentile of the sorted data is taken as the deviation threshold. Alternatively, the deviation value that best matches the assessment result with the actual accident occurrence can be selected as the deviation threshold by simulating the risk assessment results under different deviation values ​​multiple times.

[0054] Specifically, a deviation threshold of 0.03 was set. The risk evolution characteristics across spatial road segments were as follows: area K1 was more significantly affected by the synergistic effect mode and exhibited more pronounced risk fluctuations; the temporal evolution characteristics were a rapid increase in risk during the fog formation phase; and the risk evolution pattern based on historical risk factor data was an exponential change in risk when the core risk factor was in its extreme value range. The multi-factor joint fuzzy measure value was 0.656. Based on these evolutionary characteristics and patterns, a BP neural network was selected as the nonlinear computation method. The multi-factor joint fuzzy measure value was then substituted into the trained BP neural network model for initial computation. The initial risk assessment value was calculated to be 0.7. Then, the current risk factor data (atmospheric visibility 40 meters, road surface friction coefficient 0.6, traffic flow density 150 vehicles / km) were incorporated into the model in real time for the first iteration optimization. Through a phased computational submodule combined with the dynamic changes in the fog formation stage, the learning rate in the model iteration parameters was adjusted from 0.01 to 0.008. After iteration, a new initial risk assessment value of 0.66 was obtained. The deviation before and after the iteration optimization was calculated to be |0.7-0.66|=0.04. This deviation value is greater than the deviation threshold of 0.03, therefore... This process continues with a second iteration of optimization, incorporating new current risk factor data (atmospheric visibility 35 meters, road surface friction coefficient 0.58, traffic flow density 160 vehicles / km). The risk factor weight coefficients are fine-tuned by 0.02 based on the spatial distribution characteristics of area K1. After iteration, a preliminary risk assessment value of 0.64 is obtained, with a calculated deviation of |0.66-0.64|=0.02. This deviation is less than or equal to the deviation threshold of 0.03, at which point the iteration stops. Subsequently, the confidence interval for the preliminary risk assessment value is calculated, first using the iteratively optimized risk assessment value sequence (0.7, 0.66, ...). The standard deviation of the numerical distribution is calculated to be 0.026. The weight of the standard deviation is adjusted to 0.9 based on the risk volatility characteristics of the K1 zone, resulting in an adjusted standard deviation of 0.0234. At a 95% confidence level, the confidence interval is calculated as 0.64 ± 1.96 × 0.0234, i.e., [0.594, 0.686]. Finally, the preliminary risk assessment value of 0.64 after iterative optimization is combined with the confidence interval to form a quantitative dynamic risk assessment value of 0.64 ([0.594, 0.686]) with an attached confidence interval.

[0055] Understandably, the risk output layer iteratively optimizes the assessment value by incorporating current risk factor data in real time, determines the timing for stopping the iteration based on a deviation threshold, and calculates the confidence interval to output the final result. This process enables dynamic iterative correction of the assessment value, breaking through the limitations of traditional static assessments where the result is determined in a single output. The real-time incorporation of new data allows the assessment value to update dynamically with changes in foggy weather, and the addition of the confidence interval improves the reliability of the assessment result, ultimately achieving a dynamic, accurate, and credible risk assessment output.

[0056] It should be noted that: Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known structures and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0057] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of this application and form different embodiments.

[0058] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for dynamic risk assessment of heavy fog on highways, characterized in that, include: Historical traffic data of highways are collected, and historical risk factor datasets are extracted from the historical traffic data. The historical risk factor datasets are preprocessed, and a fog weather risk analysis database is constructed based on the preprocessed historical risk factor datasets. The correlation between risk factors and the probability of accident occurrence is obtained, the spatial distribution characteristics, temporal variation characteristics and dynamic variation trends of historical risk factor data are extracted, and the combined effect patterns among the historical risk factor data are identified. The evolutionary characteristics and patterns of risk are obtained based on the spatial distribution characteristics, temporal variation characteristics, dynamic variation trends, and combined action modes. Based on the aforementioned evolutionary characteristics and patterns, a risk fusion assessment framework is constructed, and the risk fusion assessment framework is divided into a risk factor input layer, a weight analysis layer, a fusion calculation layer, and a risk output layer. The current risk factor data is obtained, and the current risk factor data is processed sequentially through the risk factor input layer, the weight parsing layer, the fusion operation layer, and the risk output layer, and a quantitative dynamic risk assessment value with confidence interval is output. The risk factor input layer preprocesses the current risk factor data, and the weight parsing layer obtains the basic weights of the current risk factor data based on the historical risk factor data and the judgment matrix. The fusion operation layer is based on a fuzzy integral fusion model to obtain a multi-factor joint fuzzy measure of the current risk factor data; The risk output layer obtains and outputs the quantitative dynamic risk assessment value based on the evolutionary characteristics and patterns and the multi-factor joint fuzzy measure.

2. The method for dynamic risk assessment of heavy fog on highways according to claim 1, characterized in that, When determining the correlation between risk factors and the probability of an accident, the following should be considered: The historical risk factor dataset includes several historical risk factor data, including: atmospheric visibility, road surface friction coefficient, and traffic flow density. The first atmospheric visibility dataset is obtained by extracting the historical risk factor data corresponding to the maximum atmospheric visibility from the historical risk factor dataset, and the second atmospheric visibility dataset is obtained by extracting the historical risk factor data corresponding to the minimum atmospheric visibility from the historical risk factor dataset. The atmospheric visibility variation range is obtained by subtracting the atmospheric visibility minimum value from the maximum atmospheric visibility value. The atmospheric visibility accident probability variation range is obtained by subtracting the accident probability from the accident probability in the first atmospheric visibility dataset. The atmospheric visibility accident probability variation range is obtained by dividing the atmospheric visibility variation range by the atmospheric visibility accident probability variation range. The first road surface friction coefficient dataset is obtained by extracting the historical risk factor data corresponding to the maximum value of the road surface friction coefficient from the historical risk factor dataset, and the second road surface friction coefficient dataset is obtained by extracting the historical risk factor data corresponding to the minimum value of the road surface friction coefficient from the historical risk factor dataset. The change range of the road friction coefficient is obtained by subtracting the minimum value of the road friction coefficient from the maximum value of the road friction coefficient; the change range of the road friction coefficient accident probability is obtained by subtracting the accident probability of the second road friction coefficient dataset from the accident probability of the first road friction coefficient dataset; and the road friction coefficient correlation ratio is obtained by dividing the change range of the road friction coefficient by the change range of the road friction coefficient accident probability. The first traffic flow density dataset is obtained by extracting the historical risk factor data corresponding to the maximum traffic flow density from the historical risk factor dataset, and the second traffic flow density dataset is obtained by extracting the historical risk factor data corresponding to the minimum traffic flow density from the historical risk factor dataset. The traffic flow density change range is obtained by subtracting the minimum traffic flow density from the maximum traffic flow density. The traffic flow density accident probability change range is obtained by subtracting the accident probability from the accident probability in the first traffic flow density dataset. The traffic flow density correlation ratio is obtained by dividing the traffic flow density change range by the traffic flow density accident probability change range.

3. The method for dynamic risk assessment of heavy fog on highways according to claim 2, characterized in that, When determining the correlation between risk factors and the probability of an accident, the following should also be considered: When the atmospheric visibility correlation ratio is greater than the correlation ratio threshold, the correlation between atmospheric visibility and the probability of an accident is determined to be a strong correlation, and atmospheric visibility is marked as a core risk factor; when the atmospheric visibility correlation ratio is less than or equal to the correlation ratio threshold, the correlation between atmospheric visibility and the probability of an accident is determined to be a weak correlation, and atmospheric visibility is marked as a secondary risk factor. When the correlation ratio of the road surface friction coefficient is greater than the correlation ratio threshold, the correlation between the road surface friction coefficient and the probability of accident occurrence is determined to be a strong correlation, and the road surface friction coefficient is marked as a core risk factor; when the correlation ratio of the road surface friction coefficient is less than or equal to the correlation ratio threshold, the correlation between the road surface friction coefficient and the probability of accident occurrence is determined to be a weak correlation, and the road surface friction coefficient is marked as a secondary risk factor. When the traffic flow density correlation ratio is greater than the correlation ratio threshold, the correlation between the traffic flow density and the probability of an accident is determined to be strong, and the traffic flow density is marked as a core risk factor; when the traffic flow density correlation ratio is less than or equal to the correlation ratio threshold, the correlation between the traffic flow density and the probability of an accident is determined to be weak, and the traffic flow density is marked as a secondary risk factor.

4. The method for dynamic risk assessment of heavy fog on highways according to claim 3, characterized in that, When extracting the spatial distribution characteristics, temporal variation characteristics, and dynamic trends of historical risk factor data, and identifying the combined interaction patterns among the historical risk factor data, the process includes: The historical risk factor dataset is divided into subsets of different regions according to spatial road segments. The mean values ​​of each subset are calculated at different time points, and the changes in the mean values ​​reflect the spatial distribution characteristics. The numerical fluctuation values ​​of the same subset are calculated over a continuous period of time, and the numerical fluctuation values ​​reflect the temporal change characteristics. The numerical change curves of the historical risk factor data during the formation, development, and dissipation stages of fog are fitted using time series analysis methods, and the dynamic change trends are extracted by the slope of the numerical change curves. The core risk factor is combined with the secondary risk factor to obtain multiple risk factor combinations, and each risk factor combination contains at least two individual risk factors with different relationships. The historical risk factor data of the accidents that occurred are extracted from the historical risk factor dataset to obtain the accident dataset; The risk factor combination is extracted from the accident dataset to obtain the accident combination dataset, and the risk factor combination in the accident combination dataset is the accident risk factor combination; the risk factor combination is extracted from the historical risk factor dataset to obtain the overall combination dataset; the proportion of the accident risk factor combination in the overall combination dataset is the probability of occurrence of the risk factor combination accident; The individual risk factor is extracted from the accident dataset to obtain a single risk factor dataset, and the individual risk factor in the accident combination dataset is a single accident risk factor; the individual risk factor is extracted from the historical risk factor dataset to obtain an overall single risk factor dataset; the proportion of the individual accident risk factor in the overall single risk factor dataset is the probability of occurrence of the single risk factor accident.

5. The method for dynamic risk assessment of heavy fog on highways according to claim 4, characterized in that, When extracting the spatial distribution characteristics, temporal variation characteristics, and dynamic trends of historical risk factor data, and identifying the combined interaction patterns among the historical risk factor data, the method further includes: When the probability of occurrence of the combined risk factor event is greater than or equal to the sum of the probabilities of occurrence of each individual risk factor event, the combined action mode of the current risk factor combination is a synergistic action mode. When the probability of an event occurring in the combination of risk factors is less than the sum of the probabilities of each individual risk factor event, the combined action mode of the current risk factor combination is an antagonistic action mode.

6. The method for dynamic risk assessment of heavy fog on highways according to claim 5, characterized in that, When obtaining the evolutionary characteristics and patterns of risk based on the spatial distribution characteristics, temporal variation characteristics, dynamic change trends, and combined action modes, the following are included: Based on the spatial distribution characteristics, the correspondence between the mean values ​​of the subsets in different regions and the frequency of accidents is statistically analyzed. When the change in the mean value is greater than a first threshold, the risk in the current region is more affected by the synergistic effect mode than by the antagonistic effect mode. When the change in the mean value is less than or equal to the first threshold, the risk in the current region is more affected by the antagonistic effect mode than by the synergistic effect mode. Based on the time change characteristics, the correspondence between the numerical fluctuation values ​​of the same subset of data and the time period of the accident is statistically analyzed. When the numerical fluctuation value is greater than the fluctuation value threshold, the risk in the current period fluctuates rapidly with the change in the value of the core risk factor; when the numerical fluctuation value is less than or equal to the fluctuation value threshold, the risk in the current period remains stable. Based on the dynamic change trend and the slope of the numerical change curve, the risk change trend at different stages of fog is summarized. When the fog is in the formation stage, the risk increases with the increase of the slope; when the fog is in the development stage, the risk stabilizes as the slope approaches zero; and when the fog is in the dissipation stage, the risk decreases as the slope decreases. Based on the correlation between the numerical changes of the risk factors and the combined action patterns, the evolution law of risk with historical risk factor data is summarized. When the value of the core risk factor is in the extreme range and the core risk factor and the secondary risk factor form a synergistic action pattern, the risk changes exponentially; when the value of the core risk factor is in the normal range and the core risk factor and the secondary risk factor form the antagonistic action pattern, the risk changes linearly.

7. The method for dynamic risk assessment of heavy fog weather on highways according to claim 6, characterized in that, When constructing a risk fusion assessment framework based on the aforementioned evolutionary characteristics and patterns, the following should be included: A regionalized data processing submodule is set up in the risk fusion assessment framework, and risk factor data for different road sections are processed differently. A phased calculation submodule is set up in the risk fusion assessment framework, and the risk calculation logic is matched to the formation, development and dissipation stages of fog. A coupling effect quantification submodule is set up in the framework to quantify the interaction relationship between the core risk factor and the secondary risk factor under the synergistic and antagonistic modes.

8. The method for dynamic risk assessment of heavy fog on highways according to claim 7, characterized in that, When the weight parsing layer obtains the basic weights of the current risk factor data based on the historical risk factor data and the judgment matrix, it includes: An initial judgment matrix is ​​constructed based on the risk factor types of the current risk factor data. The rows and columns of the initial judgment matrix correspond to the risk factor types of the current risk factor data, and the matrix elements are assigned values ​​to represent the relative importance of different current risk factor data. The correlation results corresponding to the historical risk factor data and the evolution law of risk with the historical risk factor data are retrieved from the fog weather risk analysis database. The correlation results and the evolution law are used together as correction coefficients to adjust the matrix elements of the initial judgment matrix to obtain the corrected judgment matrix. The consistency ratio of the corrected judgment matrix is ​​calculated. When the consistency ratio is greater than the consistency threshold, the matrix elements of the initial judgment matrix are readjusted until the consistency ratio is less than or equal to the consistency threshold. When the consistency ratio is less than or equal to the consistency threshold, the current risk factor data is normalized. For negative factor data such as atmospheric visibility, the reciprocal standardization method is used. For positive factor data such as road friction coefficient and traffic flow density, the extreme value method is used to map to a unified numerical range. The normalized current risk factor data is substituted into the corrected judgment matrix. The maximum eigenvalue and corresponding eigenvector of the corrected judgment matrix are calculated using the eigenvalue method. The eigenvector is normalized to obtain the basic weight of the current risk factor data. The weight adjustment rules for different stages of fog are matched through the staged operation submodule, and the adjusted basic weight is passed to the fusion operation layer.

9. The method for dynamic risk assessment of heavy fog on highways according to claim 8, characterized in that, When the fusion operation layer obtains the multi-factor joint fuzzy measure of the current risk factor data based on the fuzzy integral fusion model, it includes: The coupling effect quantification submodule uses the adjusted base weight values ​​and the evolution pattern of risk with the historical risk factor data to determine the single-factor fuzzy measure value of each current risk factor data. The single-factor fuzzy measure value is positively correlated with the base weight values. The fluctuation range of the single-factor fuzzy measure value is calculated. When the fluctuation range of the measure value is greater than the second fluctuation threshold, the single-factor fuzzy measure value is smoothed. When the fluctuation range of the measure value is less than or equal to the second fluctuation threshold, a multi-factor joint fuzzy measure is calculated based on the single-factor fuzzy measure value. When calculating the multi-factor joint fuzzy measure, the synthesis rules of fuzzy integrals are combined with the synergistic and antagonistic effects to perform a weighted summation on multiple single-factor fuzzy measure values. The weight coefficients of the weighted summation are adjusted based on the combined effect mode of the risk factor combination. When the combined effect mode is the synergistic effect mode, the corresponding weight coefficient is increased; when the combined effect mode is the antagonistic effect mode, the corresponding weight coefficient is decreased. After the calculation is completed, the multi-factor joint fuzzy measure is passed to the risk output layer.

10. The method for dynamic risk assessment of heavy fog weather on highways according to claim 9, characterized in that, When the risk output layer obtains and outputs the quantitative dynamic risk assessment value based on the evolutionary characteristics and patterns and the multi-factor joint fuzzy measure, it includes: Based on the evolution characteristics of risk with spatial road segments and time dimensions, as well as the evolution law of risk with the historical risk factor data, a corresponding nonlinear calculation method is selected, and the multi-factor joint fuzzy measure is substituted into the nonlinear calculation method for initial calculation to obtain a preliminary risk assessment value. The current risk factor data is incorporated into the nonlinear calculation method in real time for iterative optimization. The iterative parameters are adjusted by combining the dynamic changes in the formation, development and dissipation stages of the fog through the staged calculation submodule, and the deviation of the preliminary risk assessment value before and after iterative optimization is calculated. When the deviation value is greater than the deviation threshold, new current risk factor data is continuously incorporated and the iteration parameters are adjusted in conjunction with the spatial distribution characteristics until the deviation value is less than or equal to the deviation threshold; when the deviation value is less than or equal to the deviation threshold, the iteration is stopped and the confidence interval of the preliminary risk assessment value is calculated. When calculating the confidence interval, the standard deviation is calculated using the numerical distribution of the iteratively optimized risk assessment value. The weight of the standard deviation is adjusted in combination with the risk fluctuation characteristics of different regions in the spatial distribution characteristics. The upper and lower limits of the confidence interval are determined based on the adjusted standard deviation. The iteratively optimized risk assessment value is combined with the confidence interval to form the quantitative dynamic risk assessment value with confidence interval and output.

Images